• Aucun résultat trouvé

Assessment of Uncertainty on a Digital Soil Map: a sensitivity analysis on the uncertainty indicators

N/A
N/A
Protected

Academic year: 2021

Partager "Assessment of Uncertainty on a Digital Soil Map: a sensitivity analysis on the uncertainty indicators"

Copied!
299
0
0

Texte intégral

(1)

HAL Id: hal-01606623

https://hal.archives-ouvertes.fr/hal-01606623

Submitted on 2 Jun 2020

Assessment of Uncertainty on a Digital Soil Map: a

sensitivity analysis on the uncertainty indicators

Philippe Lagacherie, Dominique Arrouays, Hocine Bourennane, Cécile Gomez,

Manuel Martin, Nicolas Saby

To cite this version:

Philippe Lagacherie, Dominique Arrouays, Hocine Bourennane, Cécile Gomez, Manuel Martin, et al..

Assessment of Uncertainty on a Digital Soil Map: a sensitivity analysis on the uncertainty indicators.

Pedometrics 2017, Jun 2017, Wageningen, Netherlands. 298 p. �hal-01606623�

(2)
(3)
(4)

Contents

Evaluating Use of Ground Penetrating Radar and Geostatistic Methods for Mapping Soil Cemented Horizon . . . 13

Digital soil mapping in areas of mussunungas: algoritmos comparission . . . 14

Sensing of farm and district-scale soil moisture content using a mobile cosmic ray probe (COSMOS Rover) . . . 15

Proximal sensing of soil crack networks using three-dimensional electrical resistivity to-mography . . . 16

Using digital microscopy for rapid determination of soil texture and prediction of soil organic matter . . . 17

Analysis of complementarities of different spectral analytics to sense soil properties . . . 18

Long-term diachronic series for soil carbon saturation evidence. A case study on volcanic soils of reunion island under sugarcane crops. . . 19

Concept of entropy in spatial distribution of vegetation in satellite images . . . 20

Digital Soil Mapping Method Based on the Similarity of Environmental Covariates in the Spatial Neighborhood . . . 21

Multivariate and multi-layer soil mapping using structural equation modelling. . . 22

Incorporating infrared spectroscopic data, land management, soil drainage and soil erosion observations into Bayesian framework for modelling soil erosion risk . . . 23

Comparing airborne and terrestrial laser scanning DTMs for high resolution topsoil pH modelling . . . 24

Multi-sensor data fusion for supervised land-cover classification through a Bayesian set-ting coupling multivariate smooth kernel for density estimation and geostatistical techniques . . . 25

Uncertainty in soil properties from the hydrological point of view: a call for new types of soil maps? . . . 26

Detecting soil microbial community shifts via field spectroscopy . . . 27

Using Near Infrared Spectroscopy in determining the mineralogical variations of the Lon-don Clay Formation, Whitecliff Bay, Isle of White, UK. . . 28

Standardization of world soil profile data to support global mapping and modelling . . . 29

Soil and Environment software, a tool for soil management . . . 30

(5)

Multitemporal Soil Pattern Analysis for Organic Matter Estimation on Arable Fields

using Multispectral Satellite Data. . . 40

Spectral mixing for vis-NIR diffuse reflectance spectroscopy . . . 41

Predictive mapping of the acidifying potential for acid sulfate soils . . . 42

Can soil spatial prediction models from different areas be similar? . . . 44

Thermal remote sensing for digital soil mapping. . . 45

Optimal stratification for validation of digital soil maps . . . 46

Spatial modeling of geomorphometric variables for natural hazard valuation to desertifi-cation in tropical zones. . . 47

Use of drone high resolution images to quantify soil erosion . . . 48

Predicting Scottish soil properties using X-ray powder diffraction . . . 49

Digital Soil Mapping of soil properties across GB: case studies from Scotland and England 50 A routine chemometrics approach to estimate soil organic carbon in croplands exploiting LUCAS topsoil database. . . 51

Integration of GPR measurements with sparse textural data for characterizing forest soils: an application of data fusion in southern Italy (Calabria) . . . 52

Determination of naturally occurring concentrations of trace elements in New Zealand soils 53 Mapping spatial variability of soil organic carbon, phosphorus and soil acidity in Zambia 54 A Method Research on Digital Soil Mapping Using ES-RS-GIS in Semi-arid Sandy Land: A Case Study of Horqin Left Back Banner . . . 55

Soil classification of multi-horizontal profiles using support vector machines and vis-NIR spectroscopy . . . 56

Using new sparsity genomic methods to improve soil chemometric models . . . 57

Transferring and spiking of soil spectral models between two south Indian villages . . . 58

Mapping the Impact of Zero Tillage on the Biophysical Properties of Soil . . . 59

Analysis of total carbon in soils from Itatiaia National Park: relationship with profile attributes and terrain covariates . . . 60

Proximal sensing of soil surface properties in relation to crusting, and rainfall-runoff processes: from portable to UAV-based platforms . . . 61

The spatial variability of soil’s plant-available water capacity, and its implications for site-specific management . . . 62

Evaluating recent and sub-recent magnetic impact records of air pollution by combined soil and bio-magnetic monitoring . . . 63

Ecosystem services provided by groundwater dependent wetlands in karst areas: carbon storage and sequestration . . . 64

Spatial explicit prediction of soil organic matter using a hybrid model composed of random forest and ordinary kriging . . . 65

Identifying soil management zones in a sugarcane field using proximal sensed electromag-netic induction and gamma-ray spectrometry data . . . 66

Evaluating the potential of simulated soil clay content by SoilGen2 model as soft data in Regression Kriging in sparsely sampled areas . . . 67

(6)

Assessment of soil ecosystem services at landscape scale by direct soil monitoring and

modelling. . . 73

Measuring functional pedodiversity using spectroscopic information. . . 74

Slakes: A soil aggregate stability android application . . . 75

Soil NIR-spectra and high-resolution satellite images to monitor the characteristics of active layer most related to permafrost thermal behaviour, Crater Lake CALM site, Deception Island, Marine Antarctica. . . 76

Large scale modelling of soil organic matter using DTM variables and geographically weighted regression . . . 77

Laser scanner technologies to monitoring mountain peatlands recovering . . . 78

Using a Portable XRF for Classifying Volcanic Paddy Soils of West Sumatra, Indonesia 79 Quantifying the uncertainty in a model reconstruction of a soilscape for archaeological land evaluation . . . 80

Soil hydrological classification mapping in Scotland using DSM and Random Forests . . 81

GIS-based multivariate predictive models for gully erosion susceptibility mapping in cal-careous soils . . . 82

The I4S approach to site-specific soil fertility management based on proximal soil sensing 83 Effects of Measurement Protocols and Data Mining Techniques on Soil Proxy Model Extraction: A Czech Case Study . . . 85

Soil organic carbon stocks prediction in Brazil. . . 86

Variation of soil property depth functions . . . 87

Raster sampling of three soil profiles from Wisconsin, USA . . . 88

High resolution modelling of soil organic carbon in West Greenland. . . 89

Past, present and future of physical, chemical and biological process knowledge in pedo-metrics. . . 90

Organic carbon in Swiss cropland soils 1985-2014 . . . 91

Seeing inside a pedologists head: are machine learning algorithms landscape specific? . . 92

The Pedon is at the Core of Digital Soil Morphometrics . . . 93

Validation of the 250m Soil Grids in Canada . . . 94

How universally is soil carbon increasing in New Zealand’s hill country? . . . 95

Soil texture estimation via mobile gamma-spectrometry: advanced evaluation using sup-port vector machines . . . 96

The power of Random Forest for the identification and quantification of technogenic substrates in urban soils on the basis of DRIFT spectra . . . 97

Mapping the patterns of organic matter decomposition in a high mountain environment 98 End of kriging? Or how tree-based Machine Learning Algorithms can be used to generate more accurate spatial predictions with combined geographical and feature space covariates . . . 99

Hyperspectral Imaging of Soil Cores . . . 100

High resolution estimation of peat depth using electromagnetic induction in a Scottish peatland . . . 101

(7)

Mapping the Suitable Sites for Rice Production Using Analytical Hierarchy Process and

Geographical Information System . . . 107

Soil salinity assessment through novel application of satellite thermography . . . 108

The German National Soil Inventory - Soil sampling for climate change abatement . . . 109

Prediction of soil organic carbon fractions using near infrared reflectance spectroscopy . 110 Developing Combined Soil-climate Indices for Crop Suitability Recommendations . . . . 111

Comparison of multinomial logistic regression and random forest classifiers in digital mapping of soil in western Haiti. . . 112

Pre-processing of on-the-go mapping data . . . 113

The use of proximal soil sensor data fusion and digital soil mapping for precision agriculture114 Effect of different soil compaction levels on prediction of soil properties using MIR spectra in situ . . . 115

Can organic carbon in soil cores be predicted by VNIR and MIR techniques in alpine landscape?. . . 116

Application of portable XRF and VNIR sensors for rapid assessment of soil heavy metal pollution. . . 117

Validation of the coarse-scale remotely sensed soil moisture data by using ground mea-surements with a hybrid geostatistical downscaling method . . . 118

Monitoring soil heavy metal distribution over three different time periods using trivariate linear mixed models . . . 119

Estimating soil profile attributes with proximal sensors and a spectral inference system. 120 Spatial modelling of soil carbon in Sri Lanka using sparse datasets with samples collected with different depth supports . . . 121

The Effect of Topography on Spatial Variation in Soil Health . . . 122

Application of colorimetric analysis of soils using flatbed scanners. . . 123

Gamma radiometric mapping of soil texture at field and regional scale . . . 124

Assessment of Uncertainty on a Digital Soil Map: a sensitivity analysis on the uncertainty indicators . . . 125

Predicting and Mapping Total Si over the main territory of France . . . 126

Comparison of methods to fill data gaps in soil profile databases . . . 127

Optimizing spatial sampling for multiple objectives . . . 128

Past, present and future of mathematical methods in pedometrics. . . 129

Mapping root depth soil water in sub-Saharan Africa. . . 130

Building a national (german) mid infrared database for soils. . . 131

Have extractable phosphorus and potassium contents evolved in French agricultural soils since 2004? . . . 132

Modelling the electrical conductivity of soil in the Yangtze delta in three dimensions . . 133

Sensing of soil organic carbon with portable spectrometers. . . 134

How does particle size, water and excitation time affect proximal soil sensing by X-ray fluorescence? . . . 135

(8)

Managing quantifiable uncertainty for digital land suitability assessments . . . 145

Useful applications of conditioned Latin hypercube sample for digital soil mapping . . . 146

Comparison of Multifractal parameters between binary and grayscale synthetic images . 147 Development of a stakeholder-oriented communication strategy for raising acceptance of soil protection measures . . . 148

Crossing the bridge between soil ecology and pedometrics at global scale . . . 149

Contribution to the study of erosive dynamics in the Ghézala dam watershed (Northern Tunisia) . . . 150

Pedometrics Quadrancentennial . . . 151

Improvements in spatial soil sample design efficiency . . . 152

Digital soil mapping in hilly relief area in Southeastern Brazil . . . 153

Spatial data infrastructures for handling soil data. . . 154

SoilML data exchange format and soil web services . . . 155

Uncertainty and results stability of three digital soil mapping algorithms applied to the soil cover of a farm situated on the north of Udmurt Republic, Russian Federation 156 Applying the diagnostic approach for the definition of soil functions – a pilot example on carbon sequestration and storage . . . 157

Information assessment of uncertainty of the soil’s isomorphism in pedons and elementary soil areas . . . 158

A Novel Pedometrics-econometrics Approach to Assess Soil Carbon Capability . . . 159

Modelling Pedo-Econometric carbon scores with VNIR spectroscopy . . . 160

Predicting artificially drained areas by means of a selective model ensemble . . . 161

Soil map disaggregation improved by soil-landscape relationships, area-proportional sam-pling and random forest implementation . . . 162

Digital Soil mapping Based on Airborne Gamma-Ray Imagery and Fuzzy logic : a Case of Upper Pasak Watershed, Thailand . . . 164

A 2D multifractal analysis based on detrended fluctuation analysis applied to El Pardo landscape . . . 165

Implementing Pedometrics Outside the Discipline: Context, Translation and Scalability 166 Soil-landscape controls on the impact of extreme warm and dry events on terrestrial ecosystems within continental Europe and the Mediterranean Basin . . . 167

Pedometric methods to optimize sampling and improve classification on a pilot site in the Mount Kenya region . . . 168

Comparison between random forest and partial least square regression of on-line vis-NIR spectroscopy measurements of soil total nitrogen and organic carbon . . . 169

Geochemical Signatures of Pristine Volcanic Ash and Soils from Krakatau as revealed by a Portable XRF spectrometer . . . 170

Challenges in using mid-infrared spectroscopy for the determination of soil physical, chem-ical, and biochemical properties on undisturbed soil samples . . . 171

Rapid sensing of petroleum-contaminated soils with mid infrared spectrometers . . . 172

Predicting and Mapping Topsoil Black carbon of France . . . 173

(9)

Prediction of Soil Carbon Stock in Oxisols of the Eastern Plains in Colombia by VNIR

Spectroscopy . . . 180

Estimating regression parameters in the presence of spatial and temporal correlation: A case study to quantify costs of soil constraints to the Australian grains industry . . 181

DSM online service: from the soil to the cloud . . . 182

Predicting soil organic carbon in Ap horizons in Sistan region, eastern Iran . . . 183

Digital mapping of soil salinity in eastern Iran . . . 184

Accounting for the measurement error to improve the accuracy of spatial modelling of soil carbon . . . 185

Monitoring of salt content in soil profile by hyper spectral imaging spectroscopy . . . . 186

DSM based renewal of the Hungarian Soil Spatial Data Infrastructure . . . 187

Soil, scale dependence and spatial variability: A new approach for assessing how soil variability changes with scale. . . 189

Comparison of Mid and Near Infrared Spectroscopy for Prediction of Soil Properties for a National Spatial Dataset. . . 190

Combining inventory data with ancillary datasets to predict forest soil organic carbon . 191 The Interactive Digital Soil Map of Sweden - a free web application for downscaling . . 192

Soil Microbial Diversity Across Different Agroecological Zones in New South Wales (NSW)193 Spatial modelling of landscape heterogeneity in soil moisture content with the assimilation of optical and radar remote sensing data . . . 194

Using in situ Vis-NIR combined with other sensing data to map clay content, soil organic carbon, and bulk density at the field scale . . . 195

Geophysical mapping of wetlands using DUALEM - challenges and possibilities . . . 196

Soil agrochemical monitoring – source for country-scale predictions and fertilization op-timization . . . 197

Digital soil mapping with Soil Land Inference Model (SoLIM) considering the spatial distance to soil samples . . . 198

High resolution land-use classification toward more accurate digital soil mapping of mala-gasy soils. . . 199

Are data collected to support farm management suitable for monitoring soil indicators at the national scale? . . . 200

Using combined model for soil pollution spatial analysis . . . 201

Three-dimensional mapping of soil organic carbon (SOC) based on multi-scale digital terrain analysis and data mining in Jiangxi Province, PR China. . . 202

Soil Organic Carbon Content of Central Chile and Its Projection Under Climate Change Scenarios . . . 203

Laser-induced breakdown spectroscopy (LIBS) for efficient quantitative determination of elemental plant nutrients in soils: A contribution to precision agriculture. . . 204

The use of Self Organizing Maps in hydrological modeling with SWAT . . . 205

Spectroscopy and remote sensing for assessment of peatland degradation . . . 206

(10)

Stacked generalization of statistical learners - a case study with soil iron content in Brazil215

‘spup’ – an R package for uncertainty propagation analysis in spatial environmental mod-elling. . . 216

Reflectance spectra and Land Surface Temperature integration obtained from Landsat on the soil granulometric quantification . . . 217

Visualizing Soil Landscapes . . . 218

Detection of soil microbial dynamics with high spatial and temporal resolution using infrared thermography and radiocarbon imaging . . . 219

Exploring extrapolation risks of spatial prediction models at global, continental and re-gional scales . . . 220

Proximal soil sensing:new tools for pedometricians . . . 221

Spatial variability of peat soil properties for different sampling scales . . . 222

Statistical analyses of landscape controls and vertical variability of soil organic carbon in permafrost-affected soils . . . 223

Mapping Soil Properties for achieving Soil Functions . . . 224

Refinement of soil maps of forested areas with help of geological maps, digital elevation model and remote sensing of vegetation . . . 225

Analysing spatial patterns of soil pollution profiles in floodplain exposed to historical environmental load using correlation of proportional similarity matrices with spatial matrices . . . 226

Near infrared index to assess soil texture and carbon content effects on soil hydrodynamic properties.. . . 227

Mapping subsoil ripening using Bayesian Generalized Linear Modelling. . . 228

Investigating the effect of moisture for using field-portable X-ray fluorescence spectrom-etry for 2.5D high-resolution geochemical mapping . . . 229

Desertification status mapping using recent machine language techniques . . . 230

Updating digital soil maps with new data: a case study of soil organic matter in Jiangsu, China . . . 231

Comparative examination of various uncertainty assessment approaches based on geosta-tistical approaches and machine learning algorithms . . . 232

Predicting of soil properties with geostatistical and statistical models using a stratified regular sampling grid. . . 233

Performance of a Less Expensive Radiometer for Estimating Soil Organic Carbon. . . . 234

Proximal soil sensing – steps needed to be taken from research to real-world applications 235

Spatial Variations of Soil Organic Carbon Stocks and the Related Environmental Factors in Volcanic Ash Soils in Northern Taiwan . . . 236

Estimation of current soil organic carbon stocks and evaluating the carbon sequestration rates under different management practices in arable soils of Taiwan . . . 237

High - fidelity mobile proximal soil sensing (350 -2500 nm) system . . . 238

Towards a pedogenic approach in quantitative soil sampling and element stocks estimation239

NDVI stratified sampling based on soil homogeneous areas. An application to rice crop in Babahoyo canton-Ecuador . . . 240

(11)

Comparing UAV airborne and proximal measurements of a gammaspectrometer for soil texture mapping . . . 247

Approaches for commercial Digital Soil Mapping, examples from South Africa. . . 248

What is the value of understanding? Comparing mechanistic soil formation and geosta-tistical modelling.. . . 249

Soil class mapping in the Quadrilátero Ferrífero, Brazil: a methodological approach of sampling and selection of covariates . . . 250

Digital soil classes map of Minas Gerais State, Brazil . . . 251

Modeling urbanization effect on soil functions in the New Moscow, Russia . . . 252

Estimation of Soil Profile Properties Using Field and Laboratory VNIR Spectroscopy . 253

The cost-efficiency of methods for monitoring soil organic carbon stock. . . 254

Utilizing the Legacy Soil Data of Macedonia:The Creation of the Macedonian Soil Infor-mation System and its use for digital soil mapping and assessment applications . . 255

Mapping soil properties using a non-stationary variance geostatistical model. . . 256

Modelling the mid-infrared information content of European soils . . . 257

Prediction of Soil Organic Matter Using VisNIR and PXRF Spectroscopy . . . 258

Potential of LUCAS for the development of regional-scale spectral models for the predic-tion of soil properties. . . 259

Time sequence division of high standard farmland construction based on cultivated land quality uniformity in administrative village and obstacle factors . . . 260

Is a national Vis-NIR library useful for filed scale predictions of soil type? . . . 261

Hydro-geomorphic Spatial Modelling for Multi-scale Coastal Acid Sulfate Soil Mapping 262

A VNIR penetrometer for soil profile sensing . . . 263

The Rapid Carbon Assessement Project: a modern soil carbon stock baseline for the conterminous United States . . . 264

Predicting the soil adsorption behavior of two model Persistent Organic Pollutants (POPs) to the soil solid phase based on spectral data and multivariate statistical analysis . 265

Fine-resolution mapping of soil carbon stock in Japanese forest based on machine-learning regression kriging . . . 266

A MGWRK technique for mapping soil electrical conductivity in the Heihe River Basin, northwest China . . . 267

Accounting for fieldwork costs in validation of soil maps: a comparison of design-based sampling strategies . . . 268

Identifying soil landscape units at the district scale by numerically clustering remote and proximal sensed data . . . 269

A prototype methodology for assessing within-field soil variation using digital soil map-ping, legacy soil datasets and satellite imagery to aid precision farming. . . 270

Teaching digital soil mapping as an example of contemporary environmental survey methods271

Data mining of soil color database . . . 272

Soil big data: Requirement and Potential in China . . . 273

(12)

Predictive modelling of soil properties using hyperspectral images and different multivari-ate regression techniques . . . 282

(13)
(14)

Evaluating Use of Ground Penetrating Radar and Geostatistic

Methods for Mapping Soil Cemented Horizon

Farideh Abbaszadeh Afshar – University of Jiroft, Iran Shamsollah Ayoubi – Isfahan university of technology, Iran

Annamaria Castrignano – CRA — Research Unit for Cropping Systems in Dry Environments (SCA), Bari, Italy

Depth to a root restricting (cemented) layer affects both soil moisture and nutrient availability. The knowledge of the variability of cemented layers within the soil profiles provides valuable infor-mation to decision-makers for agricultural activities. Soil surveys generally are time consuming, labor-intensive and costly, whereas geophysical and geostatistical methods offer a rapid, inexpen-sive and non-invainexpen-sive approach to mapping soil characteristics. The ground penetrating radar (GPR) is a geophysical tool that is a high-resolution electromagnetic technique used for many ap-plications, including assessment of groundwater resources, mineral exploration, archaeological, and environmental and agricultural studies. The agricultural applications may include such things as determining limited soil depth restricted by bedrock; hardpan and cemented horizon, water table depth, soil moisture content, and mapping shallow underground soil features affecting agricultural production. GPR combined with the geostatistical method could be an efficient and valid approach to provide large-scale measurements with fine-resolution data of substrate distribution. The ob-jectives of this study are to use GPR data and to explore the capability of geostatistical methods to incorporate these auxiliary variables for mapping cemented horizon in an arid region, Kerman province located in south-eastern Iran. The measurements were performed using an impulse GPR system with a center frequency of 250 MHz along ten parallel transects of 100 m length and 10 m distance between two consecutive transects. The processing of GPR data was performed with ReflexW7.0 Software. The GPR data were interpolated with geostatistical analyses were done by using the software package ISATIS. The results of the field observations on the pedons at study site showed the cemented horizon at a depth of 35-45 cm. The GPR radargram also showed a distinct interface between the two contrasting materials which usually corresponded to the occurrence of a strong reflection event. The strong reflection corresponded approximately to the top of cemented horizon. In addition the 3D soil model of GPR amplitudes signal by kriging method reproduces the quite variable depth of the cemented horizon and the extent of its lateral and vertical variation. In this study, the potential of geophysical and geostatistical to predict and mapping a cemented horizon was evaluated in an arid region. The variation of cemented horizon provided valuable information that is crucial in decision making for agricultural and engineering activities. The main advantages of GPR and geostatistical methods are the speed of obtaining data and continuous images, and rather good possibility to identify zones with markedly different properties.

(15)

Digital soil mapping in areas of mussunungas: algoritmos

com-parission

Valdemir S. Abreu – Institute of Rural Development of Amapá, Brazil Márcio R. Francelino – Federal University of Viçosa, Brazil

Elpídio I. Fernandes Filho – Federal University of Viçosa, Brazil Eliana de Souza – Federal University of Viçosa, Brazil

Eliana E.S. Santos – Federal University of Viçosa, Brazil

Mussunungas is an ecosystem of physiognomic forms ranging from grasslands to woodlands. It occurs over Spodosols originated from sandstones of the Barreiras Group from the Tertiary period, and is situated across the Atlantic Rainforest domain of southern Bahia and northern Espirito Santo States in Brazil. Areas of Mussunungas are particularly important for water infiltration dynamics and is under threat as its vegetation is not recognized as belonging to the Atlantic Rainforest Biome. This study aims at comparing classification algorithms to produce soil class maps of an area where Mussunungas occurs. The study area is the conservation unity, RPPN Rio do Brazil, located in Porto Seguro, Bahia, with an area of 1100 ha. A digital elevation model derived from an ALOS-PALSAR image of 12.5m spatial resolution was used to derive morphometric maps, to be used together with covariates such as geomorphology, euclidean distance between the points soil points and indexes derived from Landsat-8 images, CBERS-4, and Sentinel-1 imagery. A soil survey was carried out using a mixed-sample scheme combining data from a free-survey with data from regular grid. A total of 203 soil points were described collecting samples for chemical and physical analyses. R software was used for the classification using the ”caret” package. Three classifiers were tested: Extreme Gradient Boosting (xgbTree), C5.0, and Random Forest, with 10-fold cross-validation. From the six soil classes identified in the area, Haplic Lixisol are dominant, and as this occurs with inclusion of Dystric Cambisols, they were mapped together. Dystric Gleysols occurs associated to Fibric Histosols and was joined in the same map unity. Stagnic Podzols and Xanthic Ferralsols were thus each mapped as a single class. The xgbTree showed better Kappa (0.68), followed by C5.0 and Random Forest with 0.65 and 0.63, respectively. The similar performance of the classifiers showed no statistical difference. Larger confusion occurred between Xanthic Ferralsol and Haplic Lixisol. The user’s accuracy of Podzols was good, varying from 81 to 83% amongst classifiers. The morphometric covariates, elevation above sea, slope height and normalized height showed the highest importance for soil prediction. A field validation of the maps and visual interpretation of the maps overlayed with satellite images showed that the areas were well delineated, proving a good performance of all classifiers for produce a soil class map for the area of this study. We concluded that techniques of digital soil mapping can be used for mapping areas of mussunungas over Podzols.

(16)

Sensing of farm and district-scale soil moisture content using a

mobile cosmic ray probe (COSMOS Rover)

Niranjan Acharige – The University of Sydney, Australia Thomas Bishop – The University of Sydney, Australia Brad Evans – The University of Sydney, Australia Patrick Filippi – The University of Sydney, Australia Edward Jones – The University of Sydney, Australia Brendan Malone – The University of Sydney, Australia Mario Fajardo – The University of Sydney, Australia Uta Stockmann – The University of Sydney, Australia

In dryland cropping systems a grower only needs to know soil moisture at a few times during the year when a management decision needs to be made, for example at the start of the season when determining sowing and fertilizer rates. The information needs to be at the resolution of an agricultural field (or finer) and for the whole profile. Cosmic ray probes are a technology that can directly measure soil moisture at these resolutions, namely the horizontal footprint is a 150m radius and the depth of measurement is up to 0.7m. There is a now a mobile cosmic ray probe platform (COSMOS Rover) which could be used to provide a farm or district scale soil moisture map at key times in the year for growers. However, there are a number of methodological issues which need to be considered which we present in this work.

The first is a calibration issue which requires field measurements of soil moisture, soil carbon and soil lattice water content. These are prohibitively expensive for real-world applications and in this work, we compare the accuracy of the calibration between a detailed field survey as compared to using readily available data such as existing soil moisture probes and existing soil maps.

The second issue is how to interpolate the soil moisture measurements while accounting for the spatial support of measurements, and also variations in land use and other controllers of soil moisture within the measurement footprint. The measurement interval is 30 seconds so the horizontal support is 150m either side of the vehicle pathway over the travel distance of the vehicle in 30 seconds. For interpolation, we adopt an area-to-point kriging framework to account for the varying spatial support in addition to incorporating dense covariates (land use, geology) to improve our spatial predictions.

The approach is illustrated with a case study in Muttama creek catchment in eastern Australia where the dominant land uses are grazing and cropping.

keywords: Soil Moisture, COSMOS Rover, dryland cropping systems, calibration, spatial

(17)

Proximal sensing of soil crack networks using three-dimensional

electrical resistivity tomography

Jason P Ackerson – Texas A&M University, Department of Soil and Crop Science, USA Cristine L Morgan – Texas A&M University, Department of Soil and Crop Science, USA Kevin J McInnes – Texas A&M University, Department of Soil and Crop Science, USA Mark Everett – Texas A&M University Department of Geology and Geophysics, USA

Soil cracks function as primary conduits for water transport in expansive clay soils. Under-standing and predicting the impact of crack networks on the hydrology of such soils is difficult due in large part to the transient nature of crack networks. The size, position, and interconnected-ness of soils crack networks changes with seasonal wetting and drying cycles. Currently, there are limited tools for monitoring soil crack networks. In order to understand the impact of soil cracks on hydrology, new tools are needed that can monitor the spatial-temporal dynamics of soil crack networks. In this study we demonstrate the feasibility of electrical resistivity tomography (ERT) as a non-invasive tool for proximal sensing of soil crack networks monitoring. Three-dimensional ERT surveys were collected on a smectitic soil during a seasonal drying cycle. At the end of the drying cycle, soil cracks were measured directly by in-filling cracks with cement and photograph-ing cross-sections of the in-filled soil. Photographic data shows a good correlation with the ERT images, demonstrating the utility of ERT as a tool for spatial-temporal monitoring of soil crack networks.

(18)

Using digital microscopy for rapid determination of soil texture

and prediction of soil organic matter

Viacheslav Adamchuk – McGill University, Canada Asim Biswas – University of Guelph, Canada

Long Qi – South China Agricultural University, China Maxime Leclerc – McGill University, Canada

Bharath Sudarsan – Lizotte Machine Vision   , Canada Wenjun Ji – McGill University, Canada

Soil particle size analysis is an expensive analytical procedure. There are two primary methods accepted by certified soil analysis labs: hydrometer-based sedimentation and laser-based optical techniques. The first method requires substantial labour, while the second is conducted using costly equipment. This presentation describes a new system that is based on a commercially available digital microscope equipped with an array of light emitting diodes. It is deployed using a scratch resistant window in contact with measured soil. It takes just a few seconds to obtain a high quality image of dry, grounded and sieved soil that is typically used for conventional soil chemical tests. Our software uses an adopted wavelet decomposition technique to measure the percentage of particles within specified size intervals (e.g., sand, silt, and clay). Given that the level of magnification is known and the image resolution is around 2 micrometers, the instruments separate soil particles into their respected size category: clay - below 2 micrometers, sand - above 50 micrometers, and silt between clay and sand. The unique feature of this approach is that there is no need for sensor calibration and the results are expected to be comparable with those of conventional analytical soil laboratories. Based on our research, standard measurement errors range between 30 and 60 g per kg of soil, which is comparable to the observed dispersion among results obtained from different commercial soil labs. Furthermore, in addition to the image texture information component, image color was found to be suitable to predict soil organic matter content, which is another essential soil property when it comes to optimized management of soil resources. It was determined that, with proper calibration, soil organic matter content in mineral soil can be predicted with the standard error of measurements around 5-7 g per kg of soil. Based on these observations, we believe that the instrument developed can be used as an alternative to traditional soil particle size distribution techniques at a much lower cost and can preserve, or provide better measurement accuracy. Prediction of soil organic matter and in situ deployment of the instrument are additional options that could be pursued in certain environments.

(19)

Analysis of complementarities of different spectral analytics to

sense soil properties

Viacheslav Adamchuk – McGill University, Canada Wenjun Ji – McGill, Canada

Luc English – Logiag, Canada Jacques Nault – Logiag, Canada

Qianjun Gan – McGill University, Canada Ashraf Ismail – McGill University, Canada Asim Biswas – University of Guelph, Canada

To date, the cost of soil sampling and analysis constitutes the greatest limitation for adjusting site-specific management of agricultural inputs according to local needs. New methods, with minimum distortion of proximal soil sensing, are being considered to replace the laborious and expensive wet chemistry analytical techniques. Spectra-based techniques, including visible, near-infrared, and mid-infrared spectroscopy, laser-induced break down spectroscopy (LIBS), as well as machine vision are among the most favoured. Individually each of these methods has provided positive results when attempting to predict specific physical and chemical soil properties. The goal of this presentation is to discuss the complementarities of these methods in order to optimize instrumentation and analytics for use in streamlining the analysis of solid soil samples. All of the measurements were performed using 35 compressed air dried soil samples and assessed in terms of measurement reproducibility and the accuracy of prediction. The results of leave-one-out cross validation of locally established prediction models were ranked among each combination of agronomic soil properties and measurement techniques as well as their combinations.

(20)

Long-term diachronic series for soil carbon saturation evidence.

A case study on volcanic soils of reunion island under sugarcane

crops.

Myriam Allo – CIRAD, Reunion Alain Albrecht – IRD, Reunion Patrick Legier – CIRAD, Reunion Fabien Goge – CIRAD, Reunion Magali Jameux – CIRAD, Reunion Laurent Thuries – CIRAD, Reunion Pierre Todoroff – CIRAD, Reunion

The goal of the international ‘4 per mille’ initiative is to demonstrate that agricultural soils can play a crucial role for food security and climate change, and particularly in tropical areas where knowledge on soil carbon potential sequestration is still needed. The study is located at Reunion, a young tropical volcanic island in the Indian Ocean, 700km east of Madagascar, that presents a range of tropical soils comprising almost half of the 30 types recorded worldwide. Our work is focused on the drivers of soil organic carbon (SOC) content dynamics in the different types of volcanic soils under long-term sugarcane crops (more than 60% of the agricultural area). For doing so, a huge database on soil constituents has been mobilized. It was built over the last 20 years and represents more than 20 000 soil samples predominantly originated from sugarcane plots. Long-term diachronic series on SOC contents should be extracted from the database. Geolocated data combined with GIS tools allowed us to create a SOC map of Reunion and data mining tools, such as BRT, have been used to prioritize the drivers of SOC contents and evaluate the storage capacities of these young volcanic soils. In the conditions of the study, soil type was the main driver of SOC content, ahead of climate conditions and agricultural practices. Ferralsols, on the west and north coast, exhibit the lowest SOC content whereas Andisols, at higher altitudes, show the highest SOC contents for the 0-30cm layer. Long-term diachronic series showed almost constant SOC contents under sugarcane crops on the whole range of soils over time. Sugarcane cropping system produce high organic carbon inputs (residues and roots, 1.2 Mg C ha−1y−1) and agricultural practices in Reunion (mulching, one tillage every ten years on average) would maintain high SOC contents. All those considerations will suggest that soil carbon saturation is reached under sugarcane crops for all soil types. And hence, no more SOC storage is possible, but any land use change could decrease the soil organic carbon already stored in these soils. Soil carbon saturation rate, showed by long-term diachronic series, is therefore a better indicator than SOC content to develop soil carbon potential storage scenarios.

keywords: Soil organic carbon, tropical volcanic soils, long-term diachronic series, sugarcane, soil

(21)

Concept of entropy in spatial distribution of vegetation in satellite

images

Carmelo Alonso – Earth Observation Systems, Indra Sistemas S.A., Spain Felix Cid-Diaz – Grupo Sistemas Complejos, UPM, Spain

Rosa M. Benito – Universidad Politecnica de Madrid (UPM), Spain Ana Maria Tarquis – Universidad Politécnica de Madrid, Spain

The study of the dynamics of the vegetation cover is one of the most important applications of the observation of the Earth from the space and it is related to root zone soil moisture. Numerous studies have addressed the analysis of vegetation at different scales, spatial and temporal focusing on two aspects: the radiometric characterization of different types of vegetation and the spatial distribution of vegetation. The later is the result of a complex interaction between vegetation and certain environmental factors such as climate, animals and man activity. A previous step to understand the complex dynamics of the ecosystems is to be able to characterize the spatial patterns of this distribution in which soil play a major role.

In the present work we discuss the spatial distribution of vegetation based on entropy concept. However, ther are different expressions related to it: entropy dimension, relative entropy, config-uration entropy and configconfig-urational entropy per cell. Each one of them presents differences and complementary information.

In order to establish these comparisons, a multispectral image acquired on 8 August 2000 by the Ikonos satellite was selected. This satellite, in orbit from September 24-1999 at a height of 681 km, is capable of providing panchromatic images with a spatial resolution of the order of 1 m, and multispectral images, with 4 bands covering the visible region and the near infrared of the electromagnetic spectrum, with a spatial resolution of the order of 4 m. Both with a radiometric resolution of 2048 levels of gray (11 bits).

References

Alonso, C., Tarquis, A.M., Zúñiga, I. and Benito, R. Spatial and radiometric characterization of multi-spectrum satellite images through multi-fractal analysis. Nonlin. Processes Geophys., 24, 141-155, 2017.

Andraud, C., Beghdadi. A. and Lafait, A. Entropic analysis of random morphologies. Physica A, 207 (1–3), 208-212, 1994.

Murphy, B. What Does the Shannon Equation Really Mean? Pedometron, 39, 37-39, 2016. Piasecki, R. Entropic measure of spatial disorder for systems of finite-sized objects. Physica A 277 157 – 173, 2000.

Shannon, C.E. A mathematical theory of communication. Bell System Technical Journal, 27, 379 – 423, 1948.

(22)

Digital Soil Mapping Method Based on the Similarity of

Environ-mental Covariates in the Spatial Neighborhood

Yiming An – Institute of Geographic Sciences and Natural Resources Research, Chinese Academy Sciences;University of Chinese Academy of Sciences, China

Cheng-Zhi Qin – Institute of Geographic Sciences and Natural Resources Research, Chinese Academy Sciences, China

A-Xing Zhu – Institute of Geographic Sciences and Natural Resources Research, Chinese Academy Sciences;Department of Geography, University of Wisconsin-Madison, China

Lin Yang – Institute of Geographic Sciences and Natural Resources Research, Chinese Academy Sciences, China

It is a prevailing way to use soil samples to build soil-environment relationships and then to predict the soil properties of the study area. An assumption widely used for building soil-environment relationship is that those locations with similar soil-environmental conditions have similar soils. Based on this assumption, the soil property at each unvisited location can be predicted according to its environmental similarity to each individual soil sample, which is often computed based on environmental covariate values just on this locations and soil sample locations. The spatial neighborhood information is ignored in this way, although the soil at each location is also affected by the environmental condition of its surrounding area. With the assumption that the more similar the environmental conditions in the spatial neighborhood between two locations the more similar the soils are, a digital soil mapping method based on the spatial neighborhood similarity of environmental covariates was proposed in this research. The proposed method identifies the characteristic neighborhood size of each environmental covariate for each unvisited location and each soil sample and then calculate the spatial neighborhood similarity on individual environmental covariate between the unvisited location and each soil sample. The similarity on environmental condition between the unvisited location and each soil sample is computed by integrating the spatial neighborhood similarities on individual environmental covariates. The soil property value on each unvisited location is estimated to be the average of soil property values of soil samples weighted by the corresponding similarities on environmental condition. The prediction uncertainty is also provided. As a case study, the proposed method was applied to mapping soil organic matter (SOM) content (%) in the topsoil for an area (about 60 km2) in Heilongjiang province, China.

The evaluation results with an independent soil sample set showed that the proposed method got higher accuracy due to its consideration of the spatial neighborhood information during computing the similarity of environmental condition.

keywords: digital soil mapping, soil-environment relationship, similarity, spatial neighborhood,

(23)

Multivariate and multi-layer soil mapping using structural

equa-tion modelling

Marcos Esteban Angelini – INTA, Instituto de suelos, Netherlands

Gerard B. M. Heuvelink – Soil Geography and Landscape group - ESG, Wageningen University & Research, Netherlands

Bas Kempen – ISRIC World Soil Information, Netherlands

Soil properties interrelationships are rarely included in digital soil mapping (DSM) because most DSM models take a univariate approach and map soil properties individually. In some cases multivariate approaches have been used to incorporate correlations between soil properties. Cok-riging is a well-known example of such multivariate approach. A more recent approach is structural equation modelling (SEM), which has the advantage that it can incorporate pedologically-driven system interrelationships. We explored the use of SEM as a multivariate technique for DSM in a 23 000 km2study area in the Argentinian Pampas. The modelling process of SEM is driven by a

conceptual model, which is translated to a mathematical model that is calibrated with empirical data. Since soil processes operate along the soil profile, SEM may be suitable for multiple layer soil prediction. Hence, theoretical relationships between soil properties at different horizons can be included and used for prediction. The objectives of this study were: (1) to apply SEM for multi-layer and multivariate soil mapping; (2) test SEM functionality for model improvement through model suggestions; and (3) assess whether the modelled covariation among soil properties matches the covariation observed in the data. We applied SEM to simultaneously model and predict the lateral distribution of the cation exchange capacity (CEC), organic carbon (OC) and clay content of three major soil horizons, A, B and C. We used petrological relationships between these soil properties to build a conceptual model. Next, we derived a mathematical model and calibrated it using environmental covariates and soil data from 320 soil profiles. Environmental covariates included digital elevation model derivatives, MODIS products and distance to river as a proxy of parent material distribution. Cross-validation of predicted soil properties showed that the highest amount of variance explained statistics were achieved for OC (24%) and for clay (60%) of the A horizon and CEC (50%) of the B horizon. We also assessed the covariation of soil properties and demonstrated that SEM reproduces the system error variance-covariance more accurately than multiple linear regression, which is generally applied in DSM. Accurate modelling of the covaria-tion is important for stochastic simulacovaria-tion, such as required for uncertainty propagacovaria-tion analyses. We conclude that SEM can be used to predict several soil properties at multiple layers and, at the same time, provide graphical information of the relationships among system variables.

keywords: structural equation modelling, mechanistic models, soil property interrelationships,

(24)

Incorporating infrared spectroscopic data, land management, soil

drainage and soil erosion observations into Bayesian framework

for modelling soil erosion risk

Nikki Baggaley – James Hutton Institute, United Kingdom Mads Troldborg – James Hutton Institute, United Kingdom Jean Robertson – James Hutton Institute, United Kingdom

Estefania Pérez-Fernández – James Hutton Institute, United Kingdom Allan Lilly – James Hutton Institute, United Kingdom

We have developed a framework for a Bayesian Belief Network (BBN) to model soil erosion. A BBN is a graphical probabilistic model which allows for the integration of different types of data from various sources as well as incorporating expert knowledge where data are lacking and explicitly accounting for uncertainty. We are currently developing the model to integrate observations of erosion, land management and soil spectroscopy. This work builds on an inherent erosion risk assessment based on soil drainage, topography and, as in many erosion risk assessments, where erodibility is defined as a function of topsoil texture (Lilly et al., 2002). A key aspect to this work is to improve the representation of soil erodibility in the model. To do this firstly we are exploring relationships between measured soil aggregate stability, land management and soil properties. Concurrently we are using infrared (IR) spectroscopy to provide additional information on the overall soil physical and chemical composition, in particular, the composition of clays and organic matter which influence erodibility. Ultimately, we aim to develop a robust and rapid method to predict those soil characteristics which define erodibility in-field using FTIR spectroscopy as a means to monitor soil quality.

(25)

Comparing airborne and terrestrial laser scanning DTMs for high

resolution topsoil pH modelling

Andri Baltensweiler – Swiss Federal Institute for Forest, Snow and Landscape Research WSL, Switzerland

Lorenz Walthert – Swiss Federal Institute for Forest, Snow and Landscape Research WSL, Switzer-land

Terrestrial Laser Scanning (TLS) is increasingly used to create very high resolution digital terrain models (DTMs). However, little is known about the accuracy of TLS derived DTMs covering several hectares in heterogeneous environments. We investigated the accuracy of DTMs derived from TLS data and compared them to conventional, high-quality airborne laser scanning (ALS) based DTMs. Two different interpolation methods, TIN and IDW, where used to create DTMs with cell sizes ranging from 0.2 to 4 m. Furthermore, we examined the effect of the different DTM resolutions, accuracies and acquisition techniques on topsoil pH prediction models. The pH models were based on linear regression models with only terrain attributes as covariables. The study area is characterized by a complex mirco-topography which is covered by dense evergreen forest and dense ground vegetation. The pH of the topsoil varies in the study area between 3.5 and 7.0.

The results showed that up to a resolution of one meter, the TLS based DTMs were more ac-curate compared to the ALS DTMs for both interpolation methods. The accuracies of DTMs were also reflected in the performance of the pH models. Generally, the model performance decreased with decreasing resolution. However, the TIN based models with cell sizes of 0.2 m had a lower R2 than the models with 0.5 m resolution. The best soil pH model was based on a 0.5 m DTM derived from TLS. It explained 62% of the pH variance whereas the best ALS derived model explained 50% of the variance and was also based on a cell size of 0.5 m.

The pH models showed that submeter DTM resolutions are required to predict topsoil pH accurately in the study area because the pH varied over a large range within short distances. We conclude that TLS data significantly improve DTMs and pH models compared to the models based on high-quality ALS data in heterogeneous environments.

keywords: DTM, Topsoil pH modelling, Terrestrial laser scanning, Airborne LiDAR, Accuracy,

(26)

Multi-sensor data fusion for supervised land-cover classification

through a Bayesian setting coupling multivariate smooth kernel

for density estimation and geostatistical techniques

Emanuele Barca – Water Research Institute of the Italian Research Council (IRSA-CNR), Italy Annamaria Castrignanò – Consiglio per la ricerca in agricoltura e l’analisi dell’economia agraria, Research Unit for Cropping Systems in Dry Environments, Bari, Italy

Sergio Ruggieri – National Research Council of Italy, Institute for Agricultural and Forest Systems in the Mediterranean, Italy

Gabriele Buttafuoco – National Research Council of Italy, Institute for Agricultural and Forest Systems in the Mediterranean, Italy

The data fusion is a growing research field, which finds a natural application in the remote sensing, in particular, for performing supervised classifications by means of multi-sensor data. From the theoretical standpoint, to address such an issue, the Bayesian setting provides an elegant and consistent framework. Recently, a methodology has been successfully proposed incorporating a geostatistical non-parametric approach for improving the estimation of the prior probabilities in the scope of the supervised classification. In this respect, a limitation affecting the Bayes computation in the multi-sensor data is the naïve approach, which considers independent all the sensor measurements. Obviously, such hypothesis is unsustainable in practice, because different sensors can provide similar information. Therefore, an enhancement of the previous described method is proposed, introducing the smooth multivariate kernel method in the Bayes framework to furtherly improve the probability estimations. A peculiar advantage of the smooth kernel approach concerns the fact that it is inherently non-parametric and consequently overcomes the multinormality data hypotesis. A case study is presented based on the data coming from the AQUATER project.

(27)

Uncertainty in soil properties from the hydrological point of view:

a call for new types of soil maps?

Gabriele Baroni – Helmholtz Centre for Environmental Research, Germany Matthias Zink – Helmholtz Centre for Environmental Research, Germany Rohini Kumar – Helmholtz Centre for Environmental Research, Germany Luis Samaniego – Helmholtz Centre for Environmental Research, Germany Sabine Attinger – Helmholtz Centre for Environmental Research, Germany

Soil properties play an important role in modeling land surface hydrological processes. How-ever, due to the strong variability detected at all spatial scales the characterization of the soil variability remains a crucial challenge, especially over large areas. For this reason, in several stud-ies, soil parameters are inferred indirectly based on hydrological measurements (e.g., streamflow, soil moisture) to improve the predictive capability of the hydrological models. For that purpose, several approaches and strategies are available in literature. However, the characterization of the uncertainty itself has received much less attention. The aim of the study is to assess the effect of different uncertainties in soil properties on simulated hydrological states and fluxes at different spatial and temporal scales. The study is conducted based on the data collected in the Neckar catchment (Germany). The original soil map is perturbed based on three methods. These perturba-tion methods introduce the same error (variance) but with different spatial structures (correlaperturba-tion lengths) that are assumed to not be resolved in the original soil map. The generated soil properties are used as input for the distributed hydrological model mHM (www.ufz.de/mhm). The model outputs (e.g., river discharge and soil moisture) are aggregated to different spatial and temporal scales. The results show how the three perturbation methods produce different results depending on the observation scale. Streamflow acting as an integrative hydrological response of the catch-ment is affected only by the perturbation of long spatial structure. Soil moisture represents more local hydrological conditions and it is affected by the small-scale variability introduced. The study underlines the importance of a correct characterization of the uncertainty in soil properties. By that, soil maps with additional information regarding the unresolved soil spatial variability would provide a strong support to hydrological modelling applications.

(28)

Detecting soil microbial community shifts via field spectroscopy

Harm Bartholomeus – Wageningen University, Netherlands Gera Hol – NIOO, Netherlands

Reliable information on plant and soil health is important for early detection and prevention of pests, diseases or abiotic stresses. While diseases in agricultural systems can often be reliably assessed via remote sensing, it is unknown to what extent subtler changes in soil biodiversity could be detected. Plants have a multitude of positive and negative interactions with soil microorganisms which can all affect plant quality. Therefore, it seems plausible to detect shifts in soil microbial communities remotely by measuring plants response, on the leaf level or even above. We tested the hypotheses that 1) plants growing with different microbial communities will vary in leaf hy-perspectral reflectance, and 2) the spectra from plant communities can be used to derive microbial communities. We measured hyperspectral reflectance patterns of Achillea millefolium and Tri-folium repens and the whole mixed plant community with 7 plant species, growing on sterilized soils inoculated with field soil. Microbial communities varied in composition caused by serial di-lution, resulting in decreasing bacterial diversity. Largest differences in the leaf vegetation indices were found between the most diverse soils and the non-inoculated control soils. The community level measurements showed stronger treatment signals than the leaf measurements. We will dis-cuss the potential and constraints for detecting changes in soil microbial communities via plant hyperspectral reflectance.

(29)

Using Near Infrared Spectroscopy in determining the

mineralogi-cal variations of the London Clay Formation, Whitecliff Bay, Isle

of White, UK.

Ibrahim Bashar – School of Earth and Environmental Sciences, University of Portsmouth, UK, United Kingdom

Andy Gibson – School of Earth and Environmental Sciences, University of Portsmouth, UK, United Kingdom

Nick Koor – School of Earth and Environmental Sciences, University of Portsmouth, UK, United Kingdom

Andrew Gale – School of Earth and Environmental Sciences, University of Portsmouth, UK, United Kingdom

Suggested geotechnical variations within the various lithologies of the London Clay Formation had been attributed to its mineralogical variations which required its constituent minerals to be further examined. Near Infrared reflectance spectroscopy has been extensively used in the predic-tion of various soil properties, identificapredic-tion of clay minerals and related soil chemical properties. The study uses the technique of Near Infrared (NIR) spectroscopy on soil samples obtained from a section of the cliff and foreshore exposures of the London Clay Formation at Whitecliff Bay, Isle of Wight, UK. The technique helps in the identification and quantification of clay minerals, particu-larly those that are responsible for susceptibility of the Formation to expansion and shrinkage. The soil samples were measured at reflectance intervals of 1300 nm to 2500 nm using ASD LabSpec 5000 spectrometer with the resultant spectra subjected to analysis using the statistical method and The Spectral Geologist (TSG) software. Results showed that the London Clay mineralogy is dominated by clay groups such as; Kaolins, White Micas, and Smectites. The research recognizes and identifies the mineralogy within the London Clay Formation by bringing together data relat-ing to composition, mineralogy, and reflectance which helps provides valuable information on the geotechnical interpretation of the Formation using spectral techniques.

(30)

Standardization of world soil profile data to support global

map-ping and modelling

Niels Batjes – ISRIC - World Soil Information, Netherlands Eloi Ribeiro – ISRIC - World Soil Information, Netherlands Johan Leenaars – ISRIC - World Soil Information, Netherlands Ad van Oostrum – ISRIC - World Soil Information, Netherlands

Procedures for collecting, compiling, standardizing/harmonizing and subsequently providing quality-assessed world soil profile data to the international community, as developed in the frame-work of WoSIS (World Soil Information Service), are described. Harmonization, as defined by the Global Soil Partnership (GSP), involves “providing mechanisms for the collation, analysis and exchange of consistent and comparable global soil data and information”. Areas of harmonization include those related to: a) soil description, classification and mapping, b) soil analyses, c) ex-change of soil data, and d) interpretations. Seen the breadth and magnitude of the task, so far we have focused on developing and applying procedures for handling and standardizing legacy soil profile data, with special attention for the selection of soil properties considered in the Global-SoilMap specifications: organic carbon, pH, texture (sand, silt, and clay), coarse fragments (> 2 mm), cation exchange capacity, electrical conductivity, bulk density, and water retention. These properties are commonly considered in digital soil mapping and can be used to address a wide range of global issues, such as food security, combatting land degradation, and adaptation and mitigation to climate change.

The standardized data are served to the international community using two formats. Static snapshots in TXT format, with a time stamp and identifier (doi), are provided to allow for consis-tent citation purposes. For example, the ‘July 2016’ snapshot includes standardized data for some 94,000 profiles. Newly standardized data are gradually added to a dynamic version of the dataset that can be accessed ‘24/7’ using WFS connection in GIS applications (some 109,000 profiles as of March 2017); the number of measured data for each property varies between profiles and with depth. Both the static and dynamic versions are freely available at: http://www.isric.org.

Future releases of WoSIS will consider a wider range of soil properties (e.g. content of nitrogen, phosphorus and other (micro) nutrients), including data derived from soil spectrometry. Instru-mental to enhanced usability and accessibility of data managed in WoSIS will be the continued harmonization of soil property values and further standardization of soil analytical method descrip-tions. Development and testing of such procedures, in partnership with data providers, will allow for the fulfilment of (future) demands for global soil information, and enable further collation of standardized soil data shared by partners and third parties.

(31)

Soil and Environment software, a tool for soil management

Francisco Bautista – Centro de Investigaciones en Geografía Ambiental, Universidad Nacional Autónoma de México, Mexico

Angeles Gallegos – Skiu, Scientific Knowledge in Use, México, Mexico

Government agencies dedicated to environmental protection often do not have soil scientists so it is necessary to generate easy-to-use tools to interpret soil properties in the evaluation of the environmental functions of soils. The objective of this work was the elaboration of a software by means of which the properties of the profile of the soil become interpretations of the environmen-tal functions of the soils. Soil and Environment (SE) software can be used to: a) evaluate the environmental functions of soils; and b) to elaborate degradation and conservation scenarios based on the erosion process. The data required to operate the software are: thickness of horizons, bulk density, stoniness, organic carbon, cation exchange capacity and texture, and others. The envi-ronmental functions of the soil that can be evaluated with SE are: a) allowing deciding to select the best sites for housing construction considering their damping power of contaminants such as heavy metals and organic substances; b) To select the soils suitable as habitat for wild flora and fauna; c) To select sites for aquifer recharge; d) Identify soils that stores over of the organic carbon contributing to reduce climate change; e) Identify soils with archaeological importance (human history, such as ancient temples or buildings); f) To appreciate soils with geological importance (history of the land, such as soils with the bones of prehistoric animals, soils with evidence of ancient sea beds, etc.); g) Identify soils with greater aptitude for food production. The evaluation results of environmental functions and predictive models can be presented by graphs. Export of tabular and graphical information is possible as well as the spatial reference data into the GIS. Friendly interface for data input and output and database management is designed for users who do not know SQL query language.

(32)

App Soil Calculator

Francisco Bautista – Centro de Investigaciones en Geografía Ambiental, Universidad Nacional Autónoma de México, Mexico

Angeles Gallegos – Skiu, Scientific Knowledge in Use, Mexico, Mexico

The Soil Calculator app has a database to store the properties of the soil, the data that can be registered and stored in the app are: a) profile information (profile name, geographic location, weather, photography and Site observations); b) basic properties (number and name of horizons, depth, aggregate stability, volume of coarse fragments, bulk density, textural class and pH); and c) auxiliary properties (nitrogen, carbon, K2O, P2O5, cation exchange capacity, NO3, field capacity

and permanent wilting point). Soil Calculator offers aids to the user to capture the basic properties of the soil; these aids were created based on the guidelines for the description of soils in the field of FAO and the USDA. Soil Calculator allows calculations at the soil profile level for the amount of fine earth, organic carbon, nitrogen, phosphorus, potassium, cation exchange capacity and nitrates in units of weight per unit area, either per hectare or square meter. It also offers the conversion of units between the International System and the English System. Soil Calculator is a tool for professionals related to the study of soils, such as farmers, foresters, urban planners, agricultural entrepreneurs, architects, experts and non-experts in the soil science. Soil calculator is available for free download in the Play Store

(33)

Algorithms for quantitative pedology

Dylan Beaudette – USDA-NRCS, USA

Pierre Roudier – Landcare Research, New Zealand

The Algorithms for Quantitative Pedology (AQP) project was started in 2009 to organize a loosely-related set of concepts and source code on the topic of soil profile visualization, aggregation, and classification into an R package. Over the past 8 years, the project has grown into a suite of related R packages that enhance and simplify the quantitative analysis of soil profile data. Central to the AQP project is a new vocabulary of specialized functions and data structures that can accommodate the inherent complexity of soil profile information; freeing the scientist to focus on ideas rather than boilerplate data processing tasks. These functions and data structures have been extensively tested and documented, applied to projects involving hundreds of thousands of soil profiles, and deeply integrated into widely used tools such as SoilWeb. Components of the AQP project currently serve an important role in routine data analysis within the USDA-NRCS Soil Science Division. The AQP suite of R packages offer a convenient platform for bridging the gap between pedometric theory and practice.

(34)

Saskatchewan Soils: Access and improvements to soil information

Angela Bedard-Haughn – Department of Soil Science, University of Saskatchewan, Canada Ken Van Rees – Department of Soil Science, University of Saskatchewan, Canada

Murray Bentham – Department of Soil Science, University of Saskatchewan, Canada Paul Krug – Department of Soil Science, University of Saskatchewan, Canada Darrel Cerkowniak – Agriculture and Agri-Food Canada, Canada

Brandon Heung – Simon Fraser University, Canada

Kent Walters – Department of Soil Science, University of Saskatchewan, Canada

Tom Jamsrandorj – Department of Computer Science University of Saskatchewan, Canada Ralph Deters – Department of Computer Science, University of Saskatchewan, Canada

Saskatchewan, a province in western Canada, has a total area of 651,900 km2and a population

of approximately 1.2 million people, nearly half of whom live in the two largest cities. The land base is dominated by agriculture in the southern half and boreal forest in the north. The sparse population, coupled with intensive resource use, places high demand on the province’s soils with limited capacity for management oversight. At present, although most of the provincial soil survey information can be downloaded for viewing on a GIS platform, the files are not user-friendly and hence are under-utilized. Most people still rely on paper or pdf copies of the original soil surveys. In addition, most of the provincial soil maps are at a scale of 1: 100,000, which limits their usefulness for landscape modeling and precision agriculture applications.

To overcome these limitations, we have undertaken to improve the accessibility and quality of Saskatchewan’s soil information. The first phase of the project is the development of an open source, user-friendly platform for viewing and querying our soil information, with the ability to develop themed maps for several of the most commonly requested soil properties, and the option to query a full suite of properties for the province as a whole and/or for specific locations. This platform is designed to be user-friendly for an audience ranging from farmers to policy makers and for applications from education to research. It also includes the option to upload information. At this time, the focus is on facilitating the uploading of georeferenced point files (photos, publications, datasets), but the capacity for uploading new soil survey data is also in development. The latter is of particular relevance given the need for refining the resolution of our soil information. The second phase of the project involves testing and implementing digital soil mapping procedures that are feasible given the challenges we face in Saskatchewan. 1) We lack a high-resolution digital elevation model that captures the variable Prairie Pothole topography. 2) We lack sufficient point data to take full advantage of the emerging data driven techniques. The ultimate goal will be to develop DSM methodologies that can be implemented on a local to regional scale, leveraging the increasing use of drones for topographic mapping to gradually build a refined map at the 1: 10,000 scale or better. This presentation will provide an overview of our progress and challenges to date.

Références

Documents relatifs

The framework builds on a theory of infectious diseases as ecological drivers of poverty [25 –27] , and allows one to measure the impact of poul- try disease dynamics on

The results of the uncertainty propagation and of the sensitivity analysis teach us that the most influ- encing uncertain parameters are linked to the design of

The results of the uncertainty propagation and of the sensitivity analysis teach us that the most influ- encing uncertain parameters are linked to the design of

When the model was driven by local meteorological observations, and local vegetation and soil texture observations, errors in ETR estimations were caused to a

At the end of this census, the critical analysis of the existing soil maps and data showed that they were not directly utilizable for our study and that was primarily due

lame mince de l’échantillon Brèche_Min illustrant le fait que la minéralisation se développe autour d’un claste de la dolomie chocolat 2 (aspect orangé et mouchetures)

Pour le musicologue, l’étude d’une performance improvisée électroacoustique combine donc plusieurs difficultés : les difficultés liées à l’analyse d’une performance

“Biofunctool”: a framework to assess the impact of agricultural practices on soil quality based on soil functions2. Authors